The present invention relates to plasma processing systems and, more particularly, to methods and apparatus for controlling the amount of ion energy and/or the plasma density in an inductively coupled plasma processing system.

Ionized gas, or plasma, is commonly used during the processing and fabrication of semiconductor devices. For example, plasma can be used to etch or remove material from semiconductor integrated circuit wafers, and sputter or deposit material onto semiconducting, conducting or insulating surfaces. Creating a plasma for use in manufacturing or fabrication processes typically begins by introducing various process gases into a plasma chamber within a plasma reactor wherein the gases are in contact with a workpiece, such as an integrated circuit wafer. The molecules of the gases in the chamber are ionized into a plasma by radio frequency (rf) energy supplied to the plasma chamber from an external power source. During processing, the plasma and ionized particles contact the workpiece.

The rf energy applied to the plasma chamber introduces an electric field that accelerates electrons which then collide with individual gas molecules causing further production of electrons and ions. There are several ways to introduce an electric field within the plasma reactor. Two common types of plasma processing systems are capacitively coupled plasma processing systems and inductively coupled plasma processing systems.

Figure 1 is an illustration of a typical capacitively coupled plasma processing system 10, e.g. as disclosed in Microchip Fabrication, 3^rd Edition, McGraw-Hill (1997), Van Zant, pp. 267,268, for use in processing and fabrication of semiconductor devices. As shown, plasma processing system 10 includes a plasma reactor 12 having within it a plasma chamber 13. Within plasma processing chamber 13 there are two electrodes 14a and 14b which form a capacitor. Electrode 14a is coupled to ground 20 and electrode 14b is connected to receive rf energy from a power supply 16 via a matching network 18. When power supply 16 is energized the rf energy is applied to the capacitive circuit formed between electrodes 14a and 14b. If ionizable gases are provided within plasma chamber 13 then a plasma 22 is formed when the rf energy is applied.

Since plasma processing system 10 has only one power supply 16, increasing the power of the rf energy produced by power supply 16 tends to increase both the density of the plasma (i.e., plasma density) and the direct current (dc) bias at electrode 14b and wafer 24. An increase in the dc bias usually causes a corresponding increase in the potential drop across the plasma sheath 26 which increases the energy (i.e., ion energy) of the ionized particles contacting wafer 24.

Figure 2 is an illustration of a conventional inductively coupled plasma processing system 30 for processing and fabrication of semiconductor devices; the system illustrated in Figure 2 is of the type disclosed in U.S. Patents, 4,948,458 and 5,571,366. Inductively coupled plasma processing system 30 includes a plasma reactor 32 having a plasma chamber 33 therein. Unlike the plasma processing system 10 of Figure 1, inductively coupled plasma processing system 30 includes two power supplies, 34 and 36, which influence the plasma created within plasma chamber 33. Power supply 34 is configured to supply rf energy, via a match network 38, to an electrode, chuck (i.e., workpiece holder) 40, located within plasma reactor 32. The rf energy power supply 34 develops a de bias on a wafer 42 which is typically located on a top surface 44 of chuck 40.

Power supply 36 is configured to supply rf energy, via a match network 46, to a coil 48 located near plasma chamber 33. Window 50, for example a ceramic plate, separates coil 48 from plasma chamber 33. Also shown, there is typically a gas supply mechanism 52 that supplies the proper chemistry required for the manufacturing process to the plasma reactor 32. A gas exhaust mechanism 54 removes particles from within plasma chamber 33 and maintains a particular pressure within plasma chamber 33. As a result, the rf energy generated by power supply 36 creates a plasma 56 provided ionizable gases are supplied to plasma chamber 33.

The control and delivery of the rf power in a plasma discharge is of fundamental importance in plasma processing. The amount of actual power in the plasma chamber greatly affects the process conditions. Significant variance in actual power delivered to the plasma chamber can unexpectedly change the anticipated contribution of other process variable parameters such as pressure, temperature and etch rate.

As illustrated in Figures 1 and 2, the most commonly used method of obtaining a predetermined rf power within the plasma chamber is to provide a match network within the power circuit. The match network essentially transforms the impedance (capacitive or inductive reactance) of the plasma discharge into a substantially resistive load for the power supply. The power supply (or power supplies) can then be set to a predetermined power level dependent upon the desired process parameters.

By way of example, a typical match network includes variable capacitors and/or inductors as matching components (for low to high rf frequencies), and variable cavity taps or matching stubs (for use at microwave frequencies). Match networks may be adjusted manually or automatically; however, most match networks adjust automatically to changing load conditions.

In an effort to further control the amount of rf power provided to the plasma chamber, in a typical plasma processing system the output from the power supply (or power supplies) is monitored and controlled. This usually occurs at the output of the power supply itself based in part on the assumption that the power losses in the match network are negligible.

However, rf power delivered to the plasma chamber has been found to be substantially less than the rf power supply output because of unexpected losses in, for example, the match network itself. To account for losses in the match network in capacitively coupled plasma processing systems, additional sensing and controlling circuitry has been added to the power circuit. For example, United States Patents 5,175,472, 5,474,648 and 5,556,549 disclose different ways to use an rf sensor and a controller to provide an additional feedback control loop circuit that adjusts the output of the power supply to reach a desired rf power level within the plasma chamber.

Such feedback techniques have not been used in inductively coupled plasma processing systems because it has long been believed that the two power supplies are independent of each other in that the rf power supplied to the coil controls the plasma density and the rf power supplied to the chuck controls the energy of the ions contacting the wafer (i.e., by controlling the dc bias). Therefore, by having the two power supplies it was assumed that additional control over the process was inherently provided by independently setting the outputs of the two power supplies and operating them in an open loop mode (i.e., without feedback).

However, in reality the plasma density and ion energy are not truly independent since there is coupling between the power supplied at the source and the wafer chuck. This coupling is illustrated in Figure 3, for example, which is a graph of the de bias versus the rf power supplied to the wafer for various rf power settings of the rf power supplied to the coil, and various gaps (i.e., 4 or 6 cm) between the top of the plasma chamber and the surface of the wafer. The data plotted in Figure 3 were collected from a TCP9600SE processing system available from Larn Research Corporation of Fremont, California. As shown, when the rf power supplied to the chuck (i.e., bottom power) increases, the magnitude of the dc bias tends to increase. However, for a given bottom power, the dc bias which is developed also depends on the rf power supplied to the coil (i.e., TCP power), and to some extent on the gap spacing 60. This coupling between the two rf power supplies is also reflected in the energy of ions contacting the wafer. Thus, while plasma density appears to be controlled exclusively-by the TCP power supply, ion energy on the other hand, is not controlled exclusively by the bottom power supply.

Thus, there is a need for a method of and apparatus for more effectively controlling the plasma density and the ion energy in an inductively coupled plasma processing system.

According to an aspect of the present invention , there is provided apparatus for controlling an RF plasma for treating a workpiece in a plasma processing chamber including a workpiece holder having an electrode comprising a first RF source, a first match network connected between the first RF source and the electrode; the first source, first match network and electrode being constructed and arranged so RF energy is coupled from the first RF source via the first match network to the electrode, and thence to the plasma, to apply a bias voltage to the workpiece, the plasma being included in a load for the first RF source; characterised by a first sensor arrangement for sensing at least one of RF electric power, current and voltage applied to the electrode by the first RF source via the first match network , a controller arrangement arranged to be responsive to an indication of the RF parameter sensed by the first sensor arrangement for controlling at least one of RF electric power, current and voltage, that the first RF source applies to the first match network.

According to another aspect of the present invention, there is provided a method of processing a workpiece with an RF plasma processing chamber having a workpiece holder including an electrode, comprising applying RF from a first source to the electrode via a first match network so the plasma is included in a load for the first source, characterised by sensing at least one of the RF electric power, current and voltage applied to the electrode and controlling at least one of RF electric power, current and voltage the first RF source applies to the first match network in response to the sensed RF electric parameter applied to the electrode.

The foregoing and other features, aspects and advantages of the present invention will become more apparent from the following detailed description of the present invention when taken in conjunction with the accompanying drawings.

• As described, Figure 1 is a block diagram of a conventional capacitively coupled plasma processing system having a single power supply;
• As described, Figure 2 is a block diagram of a conventional inductively coupled plasma processing system having two power supplies;
• As described, Figure 3 is a graph of direct current (dc) bias level as a function of power supplied to the coil of Figure 2 and power supplied to the workpiece holder of Figure 2;
• Figure 4a is a block diagram of an improved inductively coupled plasma processing system having a feedback circuit, in accordance with one embodiment of the present invention;
• Figure 4b is a block diagram of an improved inductively coupled plasma processing system having two feedback circuits, in accordance with one embodiment of the present invention;
• Figure 5 is a block diagram of one embodiment of a feedback circuit in accordance with an embodiment of the present invention that can be used in the inductively coupled plasma processing systems in Figures 4a and 4b;
• Figure 6a is a graph of the resulting load power delivered to the plasma chamber in a conventional inductively coupled plasma processing system such as that shown in Figure 2;
• Figure 6b is a graph of load power delivered to the plasma chamber in an improved inductively coupled plasma processing system, such as that shown in Figure 4a, in accordance with one embodiment of the present invention;
• Figure 7 is a simplified schematic circuit diagram of an exemplary match network and process chamber load as seen by the power supply in a conventional inductively coupled plasma processing system such as that shown in Figure 2; and
• Figure 8 is a graph of power delivery efficiency versus load resistance corresponding to the data plotted in Figures 6a and 6b.

Plasma processes are characterized and ultimately optimized by evaluating the process results on a workpiece such as, for example, a semiconductor wafer. The process results can be characterized as a function of the plasma process parameters, e.g., rf power, plasma gas pressure, temperature, plasma chamber gap spacing, types of gases used to form the plasma, and the gas flow rates. As mentioned above, for better control of the plasma process it is preferable to control the characteristics of the rf power actually delivered to the process chamber.

The methods and apparatus of the various embodiments of the present invention provide improved control over the plasma density and/or the ion energy in an inductively coupled plasma processing system. For example, with the present invention the plasma density and the ion energy can be independently controlled by one or more feedback control circuits that essentially de-couple the rf power contributions of each of the power supplies as applied to the plasma chamber.

With this in mind, Figure 4a is a block diagram illustrating, in accordance with one embodiment of the present invention, an improved inductively coupled plasma processing system 30' which is similar to the plasma processing system shown in Figure 2 above, with the exception of an additional feedback circuit 100 in the biasing power circuit.

As shown, feedback circuit 100, coupled between chuck 40 and power supply 34 , measures at least one parameter of the rf signal applied by match network 38 to chuck 40. The parameters can include, for example, the current, the voltage, the power, an impedance, and/or phase angle associated with the power circuit and the rf signal as applied to chuck 40 and the load within plasma reactor 32. Based on one or more of these sensed parameters, for example, the deposited rf power or load impedance can be measured and/or derived.

Feedback circuit 100 can, for example, include logic and/or computational circuitry that is configured to monitor the voltage (e.g., rms voltage) of the rf signal applied to chuck 40, calculate the deposited power, and compare the deposited power to a desired power level. The results of this comparison can then be supplied, for example as a feedback control signal, to power supply 34. The feedback control signal adjusts the output of power supply 34 to match the desired power level.

Those skilled in the art will recognize that other parameters can be monitored and similarly used to adjust the rf signal applied to chuck 40. By way of example, if a user desired to maintain a fixed ion energy, while varying other process parameters such as pressure, gas flow or.TCP power, the controller can be set up to maintain a fixed rms voltage level at chuck 40. This in turn would correspond to fixing the de bias and the ion energy.

Figure 4b is a block diagram illustrating, in accordance with another embodiment of the present invention, an improved inductively coupled plasma processing system 30" which is similar to the plasma processing system shown in Figure 4a above, with the exception of a second feedback circuit 102 in the plasma generating power circuit.

As shown, feedback circuit 102, coupled between coil 48 and power supply 36, measures at least one parameter of the rf signal applied by match network 46 to coil 48. As with feedback circuit 100, the parameters can include, for example, the current, the voltage, the power, an impedance, and/or phase angle associated with the power circuit and the rf energy applied to coil 48 and the load within plasma reactor 32. Based on one or more of these sensed parameters, for example, the deposited rf power or load impedance can be calculated.

By way of example, feedback circuit 102 can include logic and/or computational circuitry for monitoring the voltage (e.g., rms voltage) of the rf signal applied to coil 48, calculating the deposited power, and comparing the deposited power to a desired.power level. The results of this comparison can then be supplied, for example as a feedback control signal, to the power supply 36. The feedback control signal adjusts the output of power supply 36 to match the desired power level.

Additionally, feedback circuits 100 and 102 can share a common computational circuit, such as a computer system running computer implemented instructions. The shared computational circuit can include data considering the status of the fabrication process as well as each of the sensed parameters from the power supplies in generating one or more of the feedback control signals.

Figure 5 is a block diagram of one embodiment of a feedback circuit 100', that can be used, for example in Figures 4a and 4b, to provide a feedback control signal to a power supply in an inductively coupled plasma processing system. Feedback circuit 100' includes a sensor 104 and a controller 106. Sensor 104 measures or detects one or more parameters of the rf energy applied to load (e.g., cables, connectors, chuck or coil/resonator, plasma, wafer, or the like, and any stray elements associated with these) and outputs a sensed parameter signal to controller 106.

In an exemplary embodiment sensor 104 is an RFZ 60 rf probe, available from Advanced Energy Corporation, Fort Collins, Colorado. Using the RFZ 60, power, voltage, current and complex impedance can be derived from low level signals representative of the vector combinations of the forward and reflected rf wave. The sensor is calibrated using various, known loads and an extrapolation algorithm (provided by Advanced Energy Corporation), and harmonics are suppressed with filters within the probe.

Controller 106 receives the sensed parameter signal from sensor 104 and supplies a feedback control signal to a power supply (e.g., power supply 34 or 36). Controller 106 also receives a desired parameter level indication that defines the desired level of operation for the power supply. For example, the desired parameter level can be a setting or a signal input by the user directly to the controller, or to another circuit/device (such as the power supply itself) that provides the information to controller 106. Controller 106 can, for example, be a computer system including a processor, a memory, and an input/output mechanism.

In the preferred embodiment, controller 106 includes an Intelligent Cable device available from Advanced Energy Corporation. The Intelligent Cable device is coupled to the RFZ 60 probe and to the power control circuit of the power supply. The Intelligent Cable completes any remaining calculations and generates the feedback control signal based on a desired parameter level as received from a power control circuit (not shown) within the power supply. Additionally, the Intelligent Cable can be coupled to a computer system, such as a personal computer (PC), to provide additional parameter monitoring and/or controlling capabilities.

Without feedback circuit 100' the power circuit operates in an essentially open loop mode wherein the desired power is set, for example, at a machine console and the power supply delivers that amount of power to the match network and load. With feedback circuit 100', however, the power circuit essentially operates in a closed loop mode to control the amount of rf power, voltage and/or current supplied by the power supply to ensure that the desired power level is actually applied to the load as measured by sensor 104.

The advantages of the various embodiments of the present invention are graphically illustrated in Figures 6a and 6b which show the test results of data collected from a Lam Research TCP 9600SE system. The data of Figure 6a were collected when the system was configured as shown in Figure 2, with the exception that an rf probe collected and operated in an open loop mode. The data of Figure 6b were collected when the system was arranged in accordance with Figures 4a and 5 and operated in a closed loop mode.

In the open loop mode, as shown in Figure 6a, the measured voltage of the delivered rf power at the chuck is a function of the outputs from power supply 34 (bias power) and power supply 36 (TCP power). The voltage generated by power supply 34 is plotted along the vertical axis and the actual load power measured by sensor 104 is plotted along the horizontal axis. In this example, the bias power ranges from 100 watts to 500 watts and the TCP power from 0 watts to 800 watts, both in 100 watt intervals.

In the closed loop mode, as shown in Figure 6b, the voltage generated by power supply 34 is again plotted along the vertical axis and the actual load power measured by sensor 104 is plotted along the horizontal axis. Again, in this example, the bias power ranges from 100 watts to 500 watts and the TCP power from 0 watts to 800 watts, both in 100 watt intervals. As shown, the delivered rf power is now substantially constant and independent of the TCP power.

With this in mind, the efficiency of the power circuits is now considered. The efficiency of power delivery to the load is of considerable interest and importance when transferring recipes among different systems. The efficiency of power transfer through the network can be defined by: $$\mathrm{Efficiency}\mathrm{=}\left({\mathrm{P}}_{\mathrm{L}}\mathrm{/}{\mathrm{P}}_{\mathrm{G}}\right)$$
where P_L is the power delivered to the load, P_G is the power supplied by the power supply (neglecting power losses in the cable and assuming that the reflected power is low). For example, both P_L and P_G can be measured by the sensor and controller and recorded at defined intervals.

Figure 7 is a simplified schematic circuit 200 of an exemplary match network and process chamber load as seen by the power supply. Circuit 200 includes a match resistance R_M 202, a match inductance 204, a match serial capacitance 206, a load impedance 208, and a current (i) 210.

Hence, the efficiency can be written as: $$\begin{array}{cc}\mathrm{Efficiency}& \mathrm{=}{\mathrm{P}}_{\mathrm{L}}\mathrm{/}\left({\mathrm{P}}_{\mathrm{L}}\mathrm{+}{\mathrm{P}}_{\mathrm{M}}\right)\hfill \\ \hspace{1em}& \mathrm{=}\left({\mathrm{i}}^{\mathrm{2}}\mathrm{\times }{\mathrm{R}}_{\mathrm{L}}\right)\mathrm{/}\left\{\mathrm{(}{\mathrm{i}}^{\mathrm{2}}\mathrm{\times }{\mathrm{R}}_{\mathrm{L}}\mathrm{)}\mathrm{+}\mathrm{(}{\mathrm{i}}^{\mathrm{2}}\mathrm{\times }{\mathrm{R}}_{\mathrm{M}}\mathrm{)}\right\}\hfill \\ \hspace{1em}& \mathrm{=}{\mathrm{R}}_{\mathrm{L}}\mathrm{/}\mathrm{(}{\mathrm{R}}_{\mathrm{L}}\mathrm{+}{\mathrm{R}}_{\mathrm{M}}\mathrm{)}\hfill \end{array}$$
where P_M is the power dissipated in the match network. If the match resistance, R_M, is constant, a plot of the efficiency obtained from equation 1 versus R_i, should have the simple form of equation 2 from which a value of R_M can be extracted. The efficiency would rise from 0 when R_L equals 0 to unity when R_L is much greater than R_M and the plot should be independent of chemistry, power, pressure, etc. and depend only on the match network and load resistances. Match networks could thus give rise to different delivered powers for the same power supply if the match networks have different internal resistances. Nonetheless, by knowing the match network efficiency and using a feedback circuit the delivered power can be more effectively controlled.

Figure 8 is a graph of power delivery efficiency versus load resistance corresponding to the data of Figures 6a and 6b. The data are closely clustered around a curve described by equation 2, above, and illustrate that the efficiency of the power supply is significantly less than unity and that efficiency varies according to the magnitude of the plasma load. Thus, in the open loop mode the real power deposited in the plasma is essentially unknown, whereas in the closed loop mode the deposited power can be set precisely to some desired value.

Moreover, with the system configuration illustrated in Figure 4b (i.e., two feedback circuits), there is the added benefit of being able to control the deposited power at the source independent, theoretically, of any loss in efficiency due to the match network, stray capacitances, inductances or other elements of the power circuit and processing system.

In accordance with one embodiment of the present invention, a method of controlling ion energy in a plasma reactor includes (a) providing rf energy from a first source to the plasma reactor, (b) creating a direct current bias with the rf energy from the first source coupled to a workpiece located within the plasma reactor, (c) supplying rf energy from a second source to the plasma reactor and striking a plasma within the plasma reactor with the energy from the second rf source, (d) detecting a parameter of the energy with an rf sensor, wherein the parameter is detected after the rf energy from the first source has been conditioned by a match network, (e) comparing the detected parameter with a desired parameter level, and (f) modifying the rf output of the first source based on the comparison such chat the level of energy of the ionized particles contacting the workpiece is controlled as a result of the magnitude of the direct current bias as established by the first rf source.

In accordance with another embodiment of the present invention, a method similar to that just described controls plasma density. Thus, the similar method includes detecting a second parameter of rf supplied by a second rf source to the plasma, comparing the second parameter to a second desired parameter level and modifying the output of the second rf source based on this second comparison such that the density of the plasma within the plasma reactor is controlled.

The methods and apparatus of the present invention are applicable to other plasma processing systems having at least two power supplies, one for generating the plasma and the other for creating a dc bias. By way of example, the present invention is applicable to Inductively Coupled Plasma (ICP), Transformer Coupled Plasma (TCP), Electroncyclotron Resonance (ECR), and helicon or helical resonating plasma processing systems.